The move allows Starcom to measure viewers of a channel that never had enough mass to show up on Nielsen's sample-based system -- the agency has been eager to ink deals with HD networks but has had no way to do that with any sort of audience guarantee.
"Those HD networks are growing very fast; the year-over-year increases are astounding," said Chris Boothe, Starcom president-chief activation officer, underscoring the need for better measurement of such networks. "HD viewing of [Discovery's] 'Planet Earth' increased the overall rating by 20%, and more and more clients are producing spots and video in HD."
The move could signal the first step in an eventual migration to census-level TV-viewing data, further enabled by a federal government mandate to complete the transition from analog to digital TV by 2009.
"When I look at the technology piece, about 50% of all TV audiences you can measure by using set-top-box data. Nielsen is now about 12,000 people in their sample, using added technology, which is very expensive," said Tracey Scheppach, VP-video innovations director at Starcom USA. "By 2009 everything will have a box. We're talking about census; forget about samples."
In the meantime, it's likely agencies will use multiple data sources to strike deals with smaller, unrated networks. Already Starcom is using the second-by-second data internally to affect planning decisions.
TNS Media Research has deals with several cable and satellite companies, including 300,000 Charter households in California and 250,000 national DirecTV households. But TNS has a history of measuring this kind of data in Europe. It has measured near-census-level set-top-box data in the U.K. using data from BSkyB. Starcom is TNS's first agency client for the data.
Quality as well as quantity
Of course, the quality of a sample, as well as its quantity, is important, and Nielsen carefully chooses its panel to be representative of the entire population. Using set-top-box data culled primarily from California will not provide as representative a sample, but Mr. Boothe says it's a matter of 300,000 viewers vs. a sample of 12,000.
"Is it perfect?" Mr. Boothe asked. "Absolutely not. But there are not a lot of things that are perfect. But it's something to use vs. nothing. Previously we didn't have any data."
Even internet audiences are not yet measured on census levels. The Interactive Advertising Bureau last month sent a letter to the major online-audience-measurement companies, Nielsen/NetRatings and ComScore, pushing for online audiences to be measured at census level.